#include <tune_quda.h>

Inheritance diagram for quda::Tunable:

Public Member Functions
	Tunable ()

virtual	~Tunable ()

virtual TuneKey	tuneKey () const =0

virtual void	apply (const cudaStream_t &stream)=0

virtual void	preTune ()

virtual void	postTune ()

virtual int	tuningIter () const

virtual std::string	paramString (const TuneParam &param) const

virtual std::string	perfString (float time) const

virtual void	initTuneParam (TuneParam &param) const

virtual void	defaultTuneParam (TuneParam &param) const

virtual bool	advanceTuneParam (TuneParam &param) const

void	checkLaunchParam (TuneParam &param)

Protected Member Functions
virtual long long	flops () const =0

virtual long long	bytes () const

virtual unsigned int	sharedBytesPerThread () const =0

virtual unsigned int	sharedBytesPerBlock (const TuneParam &param) const =0

virtual unsigned int	minThreads () const

virtual bool	tuneGridDim () const

virtual bool	tuneSharedBytes () const

virtual bool	advanceGridDim (TuneParam &param) const

virtual bool	advanceBlockDim (TuneParam &param) const

virtual bool	advanceSharedBytes (TuneParam &param) const

void	writeAuxString (const char *format,...)

Protected Attributes
char	aux [TuneKey::aux_n]

Detailed Description

Definition at line 40 of file tune_quda.h.

Constructor & Destructor Documentation

quda::Tunable::Tunable ( )

inline

Definition at line 148 of file tune_quda.h.

virtual quda::Tunable::~Tunable ( )

inlinevirtual

Definition at line 149 of file tune_quda.h.

Member Function Documentation

virtual bool quda::Tunable::advanceBlockDim ( TuneParam & param ) const

inlineprotectedvirtual

Definition at line 74 of file tune_quda.h.

virtual bool quda::Tunable::advanceGridDim ( TuneParam & param ) const

inlineprotectedvirtual

Definition at line 57 of file tune_quda.h.

virtual bool quda::Tunable::advanceSharedBytes ( TuneParam & param ) const

inlineprotectedvirtual

The goal here is to throttle the number of thread blocks per SM by over-allocating shared memory (in order to improve L2 utilization, etc.). Note that:

On Fermi, requesting greater than 16 KB will switch the cache config, so we restrict ourselves to 16 KB for now.
On GT200 and older, kernel arguments are passed via shared memory, so available space may be smaller than 16 KB. We thus request the smallest amount of dynamic shared memory that guarantees throttling to a given number of blocks, in order to allow some extra leeway.

Definition at line 113 of file tune_quda.h.

virtual bool quda::Tunable::advanceTuneParam ( TuneParam & param ) const

inlinevirtual

Definition at line 205 of file tune_quda.h.

virtual void quda::Tunable::apply ( const cudaStream_t & stream )

pure virtual

Implemented in quda::TwistCloverGamma5Cuda< cFloat, sFloat >, quda::KSLongLinkForce< Float, Result, Oprod, Gauge >, quda::TwistGamma5Cuda< sFloat >, quda::CopySpinorEx< FloatOut, FloatIn, Ns, Nc, OutOrder, InOrder, Basis, extend >, quda::ContractCuda< Float2, rFloat >, quda::ExtractGhostEx< Float, length, nDim, Order >, quda::GaugeForceCuda, quda::PackSpinor< FloatOut, FloatIn, Ns, Nc, OutOrder, InOrder, Basis >, quda::CloverCuda< sFloat, cFloat >, quda::KSForceComplete< Float, Oprod, Gauge, Mom >, quda::ShiftColorSpinorField< Output, Input >, quda::CopyGauge< FloatOut, FloatIn, length, OutOrder, InOrder, isGhost >, quda::ExtractGhost< Float, length, nDim, Order >, quda::CopyGaugeEx< FloatOut, FloatIn, length, OutOrder, InOrder >, quda::copy::CopyCuda< FloatN, N, Output, Input >, and quda::Gamma5Cuda< sFloat >.